Machine learning may be right around the corner as part of your IT infrastructure, whether you think you're invested in such a technology or not.
Google wants intelligence emanating from machine data to be as readily available as the results of a keyword search.
Now, the company is taking steps to make it happen.
It's offering machine learning interfaces with open APIs for independent or enterprise programmers to use with their applications. These include Cloud Vision API, Cloud Natural Language API, and Cloud Translation API.
The search giant is also building data analysis capabilities into common Google applications, such as Google Docs, Sheets, and Slides. As a spreadsheet accumulates device information, for example, an Explorer feature can examine its patterns and generate a chart that describes them for the period of time contained on the spreadsheet.
Google product managers demonstrated the capability at a machine learning event at Google's San Francisco offices Nov. 15.
Machine learning won't be quite as simple to execute as search, but Google is trying to dispel the notion that you'll need a platoon of data scientists before you launch your first machine learning application.
The Cloud Vision API can be used to tap into an image analysis service that can identify many entities included in the image, such as a car, a sailboat, a landmark, or an animal species. Or it can search for one face across many images. It can categorize and catalogue images, following the directions of an administrator.
Cloud Vision API can also analyze images for emotions on display, doing sentiment analysis that can be used in marketing or analysis of the feelings associated with logos in various settings.
The Cloud Vision API can be trained by a company's application for what it wishes to look for in images. It can also sharpen its results with repeated use, said Rob Craft, group product manager for the Google Cloud Machine Learning services, at the San Francisco event.
Google has reduced the price on use of Vision API by 80%, thanks in part to Google's greater reliance on its custom ASICs, known as Tensor Processing Units, Craft said. The TPUs provide a 10X improvement in performance per watt of electricity consumed, he noted.
TPU hardware plugs into the hard drive slot on a Google server. It's optimized to process data in TensorFlow data sets. TensorFlow is a software library that Google developed for handling machine intelligence data. It's made the library open source code.
The Vision API is now generally available. It's a primary example of how Google is offering services on the Google Cloud Platform that make artificial intelligence and machine learning results more accessible.
Indeed, Diane Greene, senior vice president for Google Cloud Platform, kicked off the event by saying Google will "democratize" machine learning through its simple-to-use services.
[Want to see how machine learning affects IT? Read What eBay's Machine Learning Advances Can Teach IT Professionals.]
Another example is the Google Cloud Translation API, which is now available in premium form, although it remains in beta. It will become generally available sometime in 2017.
Google employs a neural net service to perform translations of 103 languages. The process can determine the speaker's context and produce translations more likely to make sense than previous automated systems could.
"We've done a lot of science in that area," said Craft in an interview after the event. Google is reducing the price for use of the API for volume users. The company has also posted a list of prices.
Both Amazon Web Services (through its QuickSight announced a year ago) and Microsoft are also offering more AI services through their cloud services.